vllm-project / vllm-project/aibrix
[RFC] High-level AIbrix CLI to manage workloads for users who is not familiar with Kubernetes
- Dominant language
- Go
- Stars
- 5.1k
- Forks
- 694
- Avg merge
- 1d 20h
- Merged PRs (30d)
- 98
Description
### Summary
This RFC proposes the development of a high-level AIbrix CLI tool to manage workloads for users who are not familiar with Kubernetes. The AIbrix CLI will provide an intuitive interface to interact with Custom Resource Definitions (CRDs), simplifying the deployment, management, and scaling of AIbrix workloads. This tool aims to abstract the complexity of Kubernetes operations, making it easier for entry-level users to leverage the full potential of AIbrix's GenAI inference infrastructure.
### Motivation
The primary motivation behind this RFC is to lower the entry barrier for users who want to utilize AIbrix's GenAI inference capabilities but lack Kubernetes expertise. Currently, managing workloads in AIbrix requires familiarity with Kubernetes concepts and commands, which can be daunting for newcomers. By providing a user-friendly CLI, we aim to:
- Simplify the deployment and management of AIbrix workloads.
- Increase accessibility and usability for a broader audience.
- Enhance user experience by reducing the need for deep technical knowledge of Kubernetes.
- Streamline operations for users, allowing them to focus more on their AI/ML tasks rather than infrastructure management.
This CLI will improve the overall efficiency and productivity of users, fostering a more inclusive environment for AI development and deployment within AIbrix.
### Proposed Change
this is some rough proposal. no need to strictly follow.
```
aibrix deploy --template --params : Deploy a workload using a specified template.
aibrix scale --workload --replicas : Scale a workload to the desired number of replicas.
aibrix status --workload : Check the status and health of a workload.
aibrix logs --workload --tail : Retrieve the latest logs for a workload.
```
### Alternatives Considered
_No response_
Contributor guide
Assessment
This issue has not been assessed yet.